Classifier Selection for Highly Imbalanced Data Streams with Minority Driven Ensemble

摘要

The nature of analysed data may cause the difficulty of the many practical data mining tasks. This work is focusing on two of the important research topics associated with data analysis, i.e., data stream classification as well as data analysis with imbalanced class distributions. We propose the novel classification method, employing a classifier selection approach, which can update its model when new data arrives. The proposed approach has been evaluated on the basis of the computer experiments carried out on the diverse pool of the non-stationary data streams. Their results confirmed the usefulness of the proposed concept, which can outperform the state-of-art classifier selection algorithms, especially in the case of high imbalanced data streams.